Statistical Differential Analysis Methods Evaluated to Improve Aging and Chronic Disease Research
Using the right analytic tools and frameworks can make all the difference in medical research. A research team including UR CTSI’s Dongmei Li, PhD, and Zidian Xie, PhD, has determined the best method for single-cell RNA sequencing (scRNA-seq) differential gene expression (DGE) analysis within the widely used Seurat R package for scRNA-seq data analysis—statistical procedures that identify genes whose expression differs in transcriptomic data, the information critical for cell function.
Their findings have broad implications in biology research fields such as aging and chronic disease. Previously, multiple methods of DGE analysis within the Seurat framework—a widely-used approach to single-cell RNA sequencing analysis—have been applied with varying levels of accuracy and results.
In “Evaluation of statistical differential analysis methods for identification of senescent cells using single-cell transcriptomics” published in Cell Reports Methods, the research team including first author Li and last author Xie determined that the DESeq2 method performs best for DGE analysis in scRNA-seq data within the Seurat framework.
“This guidance may help reduce false discoveries in biomarker identification and improve the reliability and reproducibility of results in single-cell transcriptomic studies,” Li said.
The same dataset can produce different outcomes depending on the DGE method used, causing issues with reproducibility and accuracy in single-cell studies. Merely relying on one method because it is standard or widely used can cause problems in this research.
This study systematically assessed ten analytical methods using scRNA-seq data to evaluate performance across different sample sizes, levels of sparsity, and proportions of differentially expressed genes using a variety of metrics.
Li said that cellular senescence helps maintain tissue homeostasis, but when senescent cells accumulate, they contribute to aging and related chronic diseases.
“In translational research pipelines, cleaner biomarker discovery can improve target prioritization, validation study efficiency, and the robustness of subsequent assays built around those signatures,” Li said. “Our benchmarking study provides practical guidance for selecting more effective DGE approaches, enabling the identification of robust and reproducible senescence biomarkers that can inform downstream mechanistic investigations and candidate therapeutic interventions in aging and disease contexts.”
This research project was supported in part by the TriState SenNet of the NIH SenNet Consortium, a Common Fund initiative dedicated to building a comprehensive, multi-omic atlas of senescent cells across human tissues to advance understanding of aging biology and accelerate the development of senescence-targeted therapies.
“Through methodological benchmarking and biomarker discovery, our study contributes to this broader mission by supporting the identification of robust senescence signatures and enhancing the analytical frameworks used to map senescent cell populations,” Li said.
The findings underscored the need for improved statistical approaches to DGE analysis in scRNA-seq data.
“We highlight an important methodological gap in current single-cell workflows,” Li said. “To address this limitation, we are actively developing novel DGE frameworks that leverage deep learning algorithms to enhance FDR control while improving overall detection performance.”
Other contributions came from researchers across URochester, as well as the Roswell Park Comprehensive Cancer Center, the University of Pittsburgh, The Ohio State University, and Carnegie Mellon University.
Jonathan Raab | 3/19/2026
